Yuki — a gorgeous young Japanese woman with long straight glossy black hair, dark brown almond eyes, fair porcelain skin, and a petite slim build.
This LoRA captures Yuki's face and identity for consistent renders across poses, outfits, and scenes. Retrained on the Krea 2 V30 base (RawRealDark_v2_V30) from a fresh identity set for sharper identity lock and natural skin texture with visible pores.
Trigger word: yuki_face
Recommended strength: 0.6 (0.5–0.7 also works). Her canon long straight black hair is preserved and both caucasian and korean are negated, so she stays Japanese — negate "athletic build, tall, muscular" or her petite frame fills out.
Works best with: Krea 2 base models. Photorealistic style, natural lighting.
Example prompt:
yuki_face, a gorgeous young Japanese woman with long straight glossy black hair, dark brown almond eyes, fair porcelain skin, petite slim build, wearing a printed cotton yukata loosely belted, standing on the stepping stones of an outdoor onsen in a snowy ryokan garden, thick steam rising behind her, candid three-quarter shot, cold blue evening light
Negative prompt tips: avoid "bright" before eye colour (can cause glowing eyes).
This is a fictional character. All images are AI-generated. A matched slim-body LoRA is attached on this version for a leaner silhouette.
⭐ Enjoying Yuki? A rating + 👍 on the previews makes a real difference to an indie creator. ありがとう for the support!
## Slim-body usage
```
- Stack with the face LoRA: face @ 0.6 + slim body @ 0.8, same seed
- Use on booty/back/side shots for a leaner, flat-rear look
- Skip it on front-facing shots or if you want the curvy silhouette
- Works best on Krea 2 bases
```
## Review
⭐ Enjoying Yuki? A rating + 👍 on the previews helps an indie creator climb. ありがとう for the support!
Description
v1.1 — V30 retrain.
What changed: retrained from scratch on the Krea 2 V30 base (RawRealDark_v2_V30) using a fresh 40-image identity set for sharper identity lock and more natural skin texture with visible pores. Replaces the v1.0 blend-trained face LoRA (yuki_face_v4_blend).
Training: 16 epochs / DIM32 alpha32 on RawRealDark_v2_V30 — 40 images → 640 steps. Base: Krea 2.
Same trigger word (yuki_face) — drop-in replacement for v1.0, no prompt changes needed.
Recommended strength: 0.6 (was 0.3 on v1.0 — the V30 file is sharper and wants more).
Companion: yuki_slim_body.safetensors is attached on this version as an Enhancement LoRA (face @ 0.6 + slim body @ 0.8 on booty/back shots).





